leaf demography and phenology in amazonian rain forest: a census of 40 000 leaves of 23 tree ...

21 Pages·2004·1.14 MB·English

EcologicalMonographs,74(1),2004,pp.3–23
(cid:113)2004bytheEcologicalSocietyofAmerica
LEAF DEMOGRAPHY AND PHENOLOGY IN AMAZONIAN RAIN FOREST:
A CENSUS OF 40 000 LEAVES OF 23 TREE SPECIES
PETER B. REICH,1,6 CHRISTOPHER UHL,2 MICHAEL B. WALTERS,3 LAURA PRUGH,4 AND DAVID S. ELLSWORTH5
1Departmentof ForestResources,Universityof Minnesota,St. Paul,Minnesota55108 USA
2Departmentof Biology,PennsylvaniaState University,UniversityPark,Pennsylvania16802 USA
3Departmentof Forestry,Michigan State University,East Lansing,Michigan 48824 USA
4Departmentof Zoology,Universityof BritishColumbia,Vancouver,BritishColumbia,Canada V6T 1Z4
5School of Natural Resourcesand Environment,Universityof Michigan,Ann Arbor,Michigan48109 USA
Abstract. The periodicity, synchrony, and causes of variability in the demography of
tree leaves in ecosystems with relatively aseasonal climates, such as tropical rain forests,
is still poorly understood. To address this issue, we surveyedthetimingofbirthanddeath
of (cid:46)40000 leaves of 1445 individuals of 23 evergreen rain forest species in several late
primary and early secondary successional plant communities at San Carlos de Rio Negro,
Venezuela, in the northern Amazon basin. In all species, the mortality rate generally in-
creased with leaf age. However, in many species, deceleration of death rateswithextreme
leaf age was noted. In general, for each species, the age structure of leaf populations and
the frequency distribution of leaf life span were broad. Species differed substantially in
theirleafdemography.Measuredintheirnativehabitats,sevenspeciescommontodisturbed
open sites had shorter median life spans (0.7 yr) than five species common to open but
infertile Bana primary communities (1.9 yr average) or six species common to two tall
primary forest communities (Tierra Firme and Caatinga), when measured in high-light
conditions in the canopy (2.0 yr average).
Variation in light availability had consistent effects on leaf life span in all species.
Species native to Tierra Firme forest had average leaf life spans of 3.2, 1.9, and 1.6 yr,
respectively, in deeply shaded understory microsites, in small gaps, and in sunlit mature
tree canopies. Species nativeto Caatingaforesthadaverageleaflifespansof4.2,3.4,and
2.5 yr, respectively, in these same microsite types. Two species common in gaps and in
disturbedsiteshadmuchlongerleaflifespaninshadedunderstorythaninopen,disturbed
microsites.Forallspecies,responsesweresimilarwhentreeswereplantedinsitesdiffering
in light availability, as when trees naturally established across light gradients.
The rate of leaf production, the risk of leaf mortality, and the leaf life span were not
periodic or related consistently to seasonality of climate. Negligible relationships existed
between the mild annual dry season and either leaf production or leaf mortality in all
species. Thus, leaf phenology and demography were essentially aseasonal in this tropical
forest environment.
Key words: Amazonianrain forest; leaf lifespan; leaf longevity;light;nutrients.
INTRODUCTION chronized with seasonal patterns. A similar linkageof
tree species phenology with seasonality occurs in
Leaf longevity and phenology are important char-
many,butnotall,speciesinmarkedlyseasonaltropical
acteristics of species that reflect the influence of evo-
forests and woodlands (Frankie et al. 1974, Reichand
lution and the environment on plant traits and that, in
Borchert1984,ShuklaandRamakrishnan1984,Reich
turn,havesubstantialimplicationsforplantfunctioning
1995, Williams et al. 1997). In contrast, in relatively
atleaf,whole-plant,andecosystemscales(Chabotand
aseasonalenvironments,phenology,longevity,andde-
Hicks1982,Coley1988,Reichetal.1992,1997).This
mography may be relatively asynchronous within and
is true in tropical forests as elsewhere (Frankie et al.
amongdifferentspecies(Borchert1980),butthedegree
1974, Borchert 1980, Reich and Borchert 1984,Coley
of asynchrony and details about leaf survivorship and
1988,Reichetal.1991,Mulkeyetal.1995),although
mortality patterns are not well known.
much uncertainty remains for such systems. In the
Dataregardingtheleaflifespanoftropicalrainfor-
markedlyseasonaltemperateandborealforestbiomes,
est plants have only become available during the past
trees of most species have leaf phenologies,lifespans
quartercentury.ApioneeringstudybyBentley(1979)
(ifdeciduous),anddemographiesthatareusuallysyn-
reported two-year survivorship of 135 leaves (five
leavesfromeachof27woodyspecies).Sincethattime,
Manuscript received 17 June 2002; revised 20 November only a relatively small number ((cid:59)15–25) of detailed
2002;accepted30December2002;finalversionreceived18Feb-
ruary2003.CorrespondingEditor:S.D.Smith. leaf demography studies have been made that are rel-
6E-mail:[email protected] evanttotheecologyoftropicalevergreenmoistforests.
3
4 PETERB.REICHET AL. EcologicalMonographs
Vol.74,No.1
Manyofthepublishedstudies(e.g.,Bentley1979,Wil- ditions, both genotypic and environmental factors are
liams et al. 1989, Hegarty 1990, Miyaji et al. 1997, atplay.Althoughitisgenerallyconsideredthatshade-
Nitta and Ohsawa 1997) were carried out for a rela- tolerantspeciestendtohavegreaterleaflongevitythan
tivelyshorttime(1–2years)and/orusedindirectmeth- intolerantsinrainforests,evenundercomparablelight
ods of assessing leaf life span (Williams et al. 1989, conditions(Coley1988,Williamsetal.1989,Luskand
Sterck 1999), so in these cases little information is Contreras 1999; but see Valladares et al. 2000), there
available on seasonal, intercohort andinterannualvar- are surprisingly few adequate tests of this.
iation.Severalstudiesmadeoverperiodsof4–10years Evidence on acclimation responses to variation in
(Clark et al. 1992, Lowman 1992a,b, Sharpe 1997) light is also surprisingly scant. Treeleavesgrowingin
have addressed issues of intercohort and interannual deeplyshadedcanopypositionslivedlongerthanthose
variability.Beyondthesestudies,fewdataareavailable in high-light environments in a total of 10 temperate
on long-term patterns for multiple species and indi- evergreen, subtropical, and tropical species (Lowman
viduals within a forest. 1992a,Reichetal.1995,Miyajietal.1997).However,
Thecollectivebodyofworkreferencedabovefound in three studies with a total of 17 woody tropicalspe-
a wide variation in leaf longevity and demography cies, only four species had shorter leaf life span when
amongspecies,butsignificantrelationshipsofleafpro- growing in higher light microsites (Williams et al.
duction, mortality, and life span to seasonality (either 1989, Sterck 1999, Rose 2000). Thus, data are incon-
precipitationortemperature).Forinstance,ratesofleaf sistent regarding whether extended leaf longevity in
productionvariedwithseasonsinsubtropicalAustralia deeper shade is a common acclimation response.
(Hegarty1990,Lowman1992a),inevergreensubtrop- To address the above issues, we made a long-term
ical forest in Japan (Nitta and Ohsawa 1997), in mon- study ofthe demographics ofleavesof23treespecies
tane tropical cloud forest in Mexico (Williams-Linera growing in several different communities within an
2000), and in wet tropical forest in Panama (Coley Amazonian rain forestcomplex.Thisinvolvedalong-
1983, Aide 1993). Rates of leaf mortality (when mea- term(9-year)censusofthebirthanddeathof(cid:59)40000
sured) also were seasonal in each of those studies. In leavesof1445individualsatSanCarlosdeRioNegro,
most of the few studies where patterns were reported, Venezuela (hereafter referred to as San Carlos). We
leaf life spans varied for cohorts producedatdifferent studied naturally established and planted trees in both
times of year or for leaves that died at different times high-andlow-lightconditionsandassessedchangesin
of year (Shukla and Ramakrishnan 1984, Lowman leaf production and mortality in relation to climate,
1992a, Nitta and Ohsawa 1997). The observed varia- year,andenvironmentalposition.Therelationshipsbe-
tion in leaf life span was apparently related to the in- tweentheaveragelifespanofasmallsubsetofleaves
fluence of modest to marked seasonality(coolseasons in selected environments and other physiological and
insubtropicalforestsanddryseasonsinmoisttropical ecological processes have been addressed in previous
forests) on leaf mortality rates. reports(Uhl1987,Reichetal.1991,1994,1995).Here
Hence, where even moderate seasonality (in tem- wefocusindetailonthevariationinleafdemography,
perature ormoisture)occurs, datasuggestthatclimate including leaf production and life span, within and
exerts marked control on the temporal patterns of leaf among species and environments, to address the fol-
dynamics. The evidence is not definitive in areas with lowing set of questions and related hypotheses:
mildly seasonal climates (Cuevas and Medina 1986, 1)Howdospeciescommontodifferentcommunities
Miyajietal.1997,Sharpe1997).Therefore,thedegree and successional stages, or adapted to different light
to which tropical rain forest tree species exhibit sea- habitats, differ in the demographics of their leaf pop-
sonality and synchrony of leafdemographyisstillun- ulations?H1:Speciescommontolow-resourcehabitats
clear, and many important questions remain unan- will have longer leaf life span than those common to
swered.Howsimilarisleaflifespanamongindividuals higherresourcehabitats,evenwhenmeasuredinacom-
within species, or within individuals among years? mon environment.
Whatistheprobabilityofmortalityforaleafasitages 2) How does light environment influence variation
or in relation to modest seasonal variation in climate? in leafdemographywithinspecies?H2:Leaflifespan
Do species differing in shade tolerance vary in leaf will increase with increased shading.
demography within and across varying light environ- 3)Howdoestheriskofmortalityvarywithleafage?
ments? Do age-related changes in mortality follow similar or
Light availability affects leaf life span directly via differentpatternsamongspeciesandcommunities?H3:
plastic responses (i.e., acclimation)(AckerlyandBaz- Once beyond the juvenile stage, mortality rates will
zaz 1995, Miyaji et al. 1997) and indirectly via ad- accelerate with leaf age.
aptation in leaf life span (Coley 1988). As a result, 4) How heterogeneous is leaf production, mortality,
shade-tolerant tropical rain forest species have longer and longevity among individuals?Giventhatdifferent
leaf life span than intolerant species when both are individuals grow at different rates, in varying micro-
measured in their most common native habitat (Wil- habitats, and exhibit different degrees of suppression
liams et al. 1989, Reich et al. 1995). Under such con- fromcompetitors,itwouldnotbesurprisingifdifferent
February2004 TROPICALRAINFORESTLEAFCENSUS 5
TABLE 1. Monthly and annual rainfall, and the median rainfall (precipitation) and pan evaporation (evaporation) (all in
mm)at San Carlosde Rio Negro,Venezuela,over an 11-yearperiod.
Year Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Total
1980 132 89 310 209 306 369 401 285 489 345 365 107 3408
1981 152 248 272 369 328 382 545 276 172 166 168 308 3385
1982 133 224 277 368 492 454 520 377 217 193 231 210 3696
1983 73 201 230 329 401 359 334 247 305 187 137 219 3022
1984 283 256 217 212 425 334 372 424 240 410 249 250 3670
1985 115 64 299 169 428 413 445 419 253 235 289 176 3305
1986 220 253 179 274 296 387 442 251 254 326 347 346 3575
1987 185 156 193 322 394 363 340 340 394 217 191 100 3194
1988 304 163 101 158 331 565 459 265 345 240 311 225 3466
1989 256 150 290 254 434 404 333 216 246 381 281 173 3418
1990 232 117 274 312 456 518 406 370 107 241 202 454 3690
Medianprecipitation 185 163 272 274 401 387 406 285 253 240 249 219 3408
Medianevaporation 154 142 167 140 137 115 131 148 153 152 143 149 1703
Note:DatafromVenezuelanHydrologyand MeteorologyNetwork.
individuals showed marked variation in leaf demog- morphological positionsandsoiltypesthatoccurwith
raphy in an aseasonal climate (Borchert 1980). H4: slightelevationdifferences(CuevasandMedina1986,
Phenology will be asynchronous among individuals Reich et al. 1991). We studied 15 species from four
and species. undisturbedlate-successionalforestcommunitieswith-
5)Doesseasonalityinfluencetheratesofleafbirths, in 1 km of each other: species-rich Tierra Firme on
deaths, and overall patterns of leaf longevity or phe- oxisol,legume-dominatedTierraFirmeonultisol,Tall
nology?Istheriskofmortalitygreaterduringthe(rel- Caatinga, and Bana. Tierra Firme forests occur on the
atively mild) annual dry season at San Carlos, and if highestsites,whiletheCaatingaandBanacommunities
so, does this differ among species and/or among leaf areatlowerelevation,onperiodicallyfloodedfine-sand
ageclasses(i.e.,young,mature,oldleaves)?H5:Leaf andcoarse-sandsites,respectively(CuevasandMedina
production will be lower and mortality rates will be 1986). The Bana community is considered the most
higher during the annual dry season than during the nutrientlimitedofthetreeprimarycommunities(Cue-
wet season, and in dry years vs. wet years. vasandMedina1986,Reichetal.1994).TheCaatinga
community is more N limited than the Tierra Firme
METHODS community,whichmaybemorePorCalimited(Reich
Sites and species etal.1995).Inaddition,westudiedeightspeciesgrow-
TheresearchsitewaslocatednearSanCarlosdeRio ing in cultivatedand/orrecentlyabandonedfarmplots
Negro, Venezuela (1(cid:56)56(cid:57) N, 67(cid:56)03(cid:57) W) at 119 m ele- on Tierra Firme sites. We studied cultivated Manihot
vationinthenorth-centralAmazonBasin.Theclimate esculenta in farm plots, and seven early-successional
of the region is humid equatorial, with a mean tem- treespeciesthatcolonizefarmplotsduringcultivation
perature of 26(cid:56)C, a mean annual rainfall of 3565 mm, and after abandonment. Sites were cultivated for 1–2
and a mean annual potential evaporative demand of yearsfollowingforestcuttingandburning.Throughout
1700–1900 mm (Ministerio del Ambiente y de Recur- this paper we will refer to species by genera, unless
sosNaturalesRenovables).Temperaturevariesslightly genus and species names are required for clarity.
month-to-monthoryear-to-year;themeantemperature Thelegume-dominatedTierraFirmecommunitywas
is between 25(cid:56) and 27(cid:56)C in every year. During the representedbyonlythedominanttreespecies,Eperua
period of this study ((cid:59)1982–1990), annual precipita- purpurea. Data for this species are lumped hereafter
tion ranged from a low of 3022 mm in 1983 to a high with the other Tierra Firme species. Each of the other
of 3696 mm in 1982 (Table 1). Although substantial threeprimarycommunitieswasrepresentedbyfouror
rainfalloccursthroughouttheyear,thereispronounced five common species. Bana trees are relatively short
seasonality, with December through March being the and open grown, and a leaf demography census was
driestperiodandMaythroughJulythewettesttimeof carried out for midcanopy branches on such trees.For
year.Thedriestmonthsinthe9-yearstudyperiod(and Tierra Firme and Caatinga (both tall forests), we cen-
theonlymonthswith(cid:44)100mmrainfall)wereFebruary susednaturallyestablishedindividualsofallspeciesin
1985 (64 mm) and January 1983 (73 mm). In both of deeplyshadedforestunderstory;forasubsetofspecies,
these months, pan evaporation exceeded precipitation wecensusedleavesofindividualsinforesttreefallgaps
by (cid:46)100 mm. The driest 2-month period during this and in canopy branches of tall, dominant trees (Table
study wasJanuary–February1985whenatotalof179 2).Thesethreemicroenvironmentsrepresentagradient
mm rain fell. from deep shade to high light availability.
Severalwell-differentiatedcommunitiesoccurinthe ForfouroftheTierraFirmeandthreeoftheCaatinga
Rio Negro region, each associated with distinct geo- species,wealsocensusedbranchesofyoungpole-sized
6 PETERB.REICHET AL. EcologicalMonographs
Vol.74,No.1
TABLE2. Numberoftrees(totalnumberofleaves)andmedianandmaximumleaflifespan(notavailableinallcases)for
speciesestablishednaturallyand plantedin variousenvironmentsat San Carlosde Rio Negro,Venezuela.
No. trees Medianleaf Maximum
Environmentand species (no. leaves) lifespan (d) lifespan (d)
TierraFirme
Farmplotopen
Caesalpinaceae,Eperuapurpurea 14 [phone]
Chrysobalanaceae,Licania heteromorpha 15 [phone]
Lauraceae,Ocoteacostulata 9 [phone]
Burseraceae,Protiumspp. 29 (1131) 687 1398
Canopy
Caesalpinaceae,Eperuapurpurea 31 [phone]
Chrysobalanceae,Licaniaheteromorpha 152 (1796) 736 1763
Lauraceae,Ocoteacostulata 138 (1594) 462 1490(cid:49)
Gap
Lauraceae,Ocoteacostulata 6 [phone]
Understory
Caesalpinaceae,Eperuapurpurea 22 (212) 1037 2341
Chrysobalanceae,Licaniaheteromorpha 31 (567) 1161 2675
Lauraceae,Ocoteacostulata 30 (378) 1122 2584
Burseraceae,Protiumspp. 30 (378) 1006 2219
Fabaceae,‘‘Cabari’’sp. 24 (165) 1544 2554
Second-growthunderstory
Chrysobalanceae,Licaniaheteromorpha 6 (429) 1502 2310
Lauraceae,Ocoteacostulata 7 (132) 1110 2037
Burseraceae,Protiumspp. 10 (1273) 1420 2219
Caatinga
Farmplotopen
Clusiaceae,Caraipa heterocarpa 10 [phone](cid:49)
Caesalpinaceae,Eperuapurpurea 8 (90) 757 1763(cid:49)
Euphorbiaceae,Micrandra sprucei 7 [phone]
Canopy
Clusiaceae,Caraipa heterocarpa 137 (2533) 885 2371(cid:49)
Caesalpinaceae,Eperuapurpurea 119 (602) 1076 1824
Euphorbiaceae,Micrandra sprucei 90 [phone](cid:49)
Gap
Clusiaceae,Caraipa heterocarpa 1 1146
Caesalpinaceae,Eperuapurpurea 7 (132) 1228 1702
Euphorbiaceae,Micrandra sprucei 8 (225) 1338 1763
Burseraceae,Protiumspp. 5 529
Understory
Clusiaceae,Caraipa heterocarpa 24 (530) 1268 2493(cid:49)
Caesalpinaceae,Eperuapurpurea 29 (188) 1389 2797
Euphorbiaceae,Micrandra sprucei 33 (401) 1623 3070(cid:49)
Sapotaceae,Micropholismaguirei 27 (413) 1620 2888(cid:49)
Burseraceae,Protiumspp. 31 (363) 1693 2797
Bana
Canopy
Apocynaceae,Aspidospermaalbum 28 [phone]
Nyctaginaceae,Neeaobovata 54 (1646) 377 1642
Burseraceae,Protiumspp. 32 [phone]
Rubiaceae,Retiniphyllumtruncatum 31 (4751) 474 1672
Bombacaceae,Rhodognaphalopsishumilis 20 [phone]
Earlysecondarysuccessional
Farmplot(open)
Euphorbiaceae,Manihot esculenta 2 (156) 55
Earlysecondgrowth(open)
Melastomataceae
Belluciagrossularioides 46 (4734) 173 1003
Clidemiasericea 24 (1084) 258 730
Miconia dispar 2 (49) 334
February2004 TROPICALRAINFORESTLEAFCENSUS 7
TABLE 2. Continued.
No. trees Medianleaf Maximum
Environmentand species (no. leaves) lifespan (d) lifespan (d)
Moraceae,Cecropia ficifolia 19 (1188) 76
Celastraceae,Goupia glabra 23 [phone](cid:49)
Clusiaceae
Vismiajapurensis 27 (4908) 255 790
Vismialauriformis 31 (2668) 234 730
Tierrafirmeunderstory
Melastomataceae
Belluciagrossularioides 24 [phone]
Miconia dispar 1 (47) 553 912
treesplantedinhigh-lightenvironmentsinrecentclear- Theleaflifespansreportedinthispaperareinsome
ings. Leaf demography was also studied for three of cases shorter than in our earlier publications(Reichet
the Tierra Firme species that had been planted in the al. 1991, 1995), for two reasons. First, the data pre-
understory of a young (5–10 year old), naturally re- sentedintheearlierleafphysiologypapersrepresented
generated,second-growthforest.Twooftheearlysuc- a small subset of all leaves, and the average life span
cessional species were studied in both high-light sites reported was pooled across light environments that
(abandoned farm plots) and in the forest understory were skewed toward the understory. Second, in those
wherethey werenaturallyestablished.Theotherearly papers (and done similarly by Lowman 1992a,b), we
successionalandBanaspecieswerestudiedinasingle used mean life span of leaves that did not die during
(open) environment. the first one-fifth of the average lifespan(considering
thatthosethatdiedyoungwerelikelyduetoherbivory,
Leaf census approach and methods disease, or other noninternally regulated factors, and
thusdidnotreflecttheintrinsiclifespan).Hencethose
Leaf demography was surveyed beginning in 1982
earlier papers used the life span of all leaves that did
on one or two branches of 1445 trees of 23 species in
not die a premature death, given the desire to relate
45 species–site combinations (Table 2). Branches in
the intrinsic leaf life span with other leaf ecophysio-
mid-to-uppercanopypositionsinlargetreeswerecen-
logical traits of young, mature leaves. Thus, using the
susedbytreeclimbing.ForallspeciesexceptManihot
entire population, leaf life spans reported here would
and Cecropia (see next paragraph),thetimingofbirth
be shorter than those reported earlier.
and death of individual leaves was tracked over time
at repeated census intervals for every tree. Census in-
Statistics
tervals typically ranged from 1 to 4 months among
species, with shorter intervals used for species with Weanalyzedforenvironmentandspeciesdifferences
fasterleafturnoverrates.Newlyproducedleaveswere in survival with proportional hazards modeling (Fox
identified as those that had emerged since the prior 1993) using JMP software (SAS 1997). Leaves that
census. A leaf (from the prior live-leaf census popu- wereharvestedforphysiologymeasurementsforother
lation) was identified as newly dead if it had either studies(e.g.,Reichetal.1991)were‘‘rightcensored’’
disappearedorwaspresentbutvisiblydead.Asystem- in the data set (this amounted to a tiny fraction of the
atic drawing of branch, subbranch, and leaf position total),aswereleavesthatlivedbeyondthefinalcensus
wasusedtomakethesesurveys,withdifferentcolored date. Right censoring indicates that a leaf wasaliveat
pencils used at every census date to draw every new thetimeitwasremovedfromthepopulationpoolused
leaf (or to indicate that no new leaves wereobserved) for statistical analysis. We chose semiparametric pro-
and to cross out every newly dead leaf (death). The portional hazards modeling over parametricapproach-
life span of each leaf was calculated as the time be- es, since preliminary analyses of Weibull, and log-log
tweenthecensusoffirstappearanceandthefirstcensus transformations of product limit (Kaplan-Meier) sur-
of death or disappearance. vival curves indicated that the risk ofmortalityvaried
For Manihot and Cecropia the mean leaf life span substantially among environments and species. Under
per plant was assessed using a technique commonly suchasituation,Fox(1993)recommendedproportional
used for species with relatively continuous and equal hazard modeling. Based on the results of the propor-
leafbirthanddeathrates(Williamsetal.1989,Ackerly tionalhazardsanalyses,wedevelopedestimatesofpro-
1999).Thetotalcumulativenumberofleavesproduced portional survival over time or leaf age for species in
and dead are plotted over time, and the mean distance each environment using the Kaplan-Meier approach.
in time between the two theoretically parallel lines is We tested for significant differences in Kaplan-Meier
equivalent to the average leaf longevity. estimates among species, within and across environ-
8 PETERB.REICHET AL. EcologicalMonographs
Vol.74,No.1
FIG.1. Mortalitycurves(i.e.,Kaplan-Meiersurvivorshipcurve)forspeciesinopenandshadedpostagriculturalsecondary
successionalTierraFirmesites(panelsa,d,e,andh)andthreedifferentprimaryforests(panelsb,c,f,g,andi),corresponding
largelytothenongapsitesinTable2.Inadditiontospeciesnaturallyregeneratedinthesesites,thesepanelsincludespecies
common to the Tierra Firme and Caatinga forests that were planted in postagricultural Tierra Firme sites with differing
degrees of canopy closure (a, d, and e). All curves were significantly different. For each curve, data from all individuals
and allleafcohortswerepooled.
ments, and within species among environments; and allcommoncensusintervals.Weusedthisapproachto
for different cohorts with log-rank and Wilcoxontests assess correlations of production and mortalityamong
(KalbfleishandPrentice1980).Thelog-ranktestplaces and within species, microenvironments, and sites. To
moreweightonlargersurvivaltimesandismostuseful do this, we classified leaves into age classes (e.g., ex-
when the ratio of hazard functions (hazard function(cid:53) panding, young fully expanded, mature, old).
instantaneousfailurerateatagiventime)inthegroups
being compared is approximately constant. The Wil- RESULTS
coxon test places more weight on early survivaltimes
Variation among species and environments
and is the optimum rank test if the error distribution
is logistic. Median leaf life span was calculatedasthe Species differed substantially within and among
half-life from the Kaplan-Meier estimates. Contrasts communities in the dynamics of their leaf population
amongcommunitytypesandhabitats(understory,gap, demography. For the 23 species (in a total of 47 spe-
canopy, farm plot) were made using analysis of vari- cies–microsite combinations), median leaf life span
ance, with species as replicates. varied from 0.2 to 4.7 yr and the maximum life span
To assess whether temporal patterns of leaf produc- was (cid:46)8.4 yr. All species common to disturbed early-
tionormortalitydisplayedperiodicityorsynchronicity successional sites tended to have short average and
among leaves of different ages, among trees within a maximum leaf life spans, while those occupying in-
species or among species on a site, we utilizedsimple fertile Bana or shaded tall primary forest sites tended
correlationsofeitherproportionalproductivityratesor to have longer leaf life spans (Figs. 1 and 2, Table 2;
percentage mortality rates per month (30.4 d) for a differences significant at P (cid:44) 0.001).
number of census dates. For example, to test whether Plantsthatpassasignificantpartoftheirlifehistory
leaf mortality in trees A and B were synchronized or inshadedunderstorieshadlongerleaflifespans(under
not, we correlated pairs of proportional mortalitydata comparablelightlevels)thanspeciesthattypicallyoc-
for all census intervals in common. In essence, if tree cupy higher light microhabitats. In the shaded Tierra
A had high proportional mortality during census in- Firme understory (Table 2), late-successional shade-
tervals where tree B also had high mortality, and low tolerant species had greater leaf life spans (averageof
mortalitywhentreeBhadlowmortality,theirmortality 3.2 yr) than Bellucia and Miconia species (1.5–1.75
rates would tend to be significantly correlated across yr),whicharemorecommonlygapinhabitants.Similar
February2004 TROPICALRAINFORESTLEAFCENSUS 9
FIG.2. Measuredvaluesforcumulativenumberofproduced(solidcircles)andsenesced(opencircles)leavesovertime
fortwoindividualseachofManihotesculentaandCecropiaficifolia(totalnumbersoftreesofeachwere2and19,respectively).
Thesedatain totalwereused in determiningthe leaflifespans in Table2.
patterns occurred at high light. Species that typically trolling for light environment, species from the more
occur early during succession under high light avail- nutrient-poor community would have longer lived
ability (Reich et al. 1995, Ellsworth and Reich 1996) leaves. The results also strongly support Hypothesis 2
have a shorter leaf life span (mean of 0.8 yr for seven that leaf longevity decreases with increasing habitat
species) than later successional species, when both irradiance. For both forest community types, trees in
groups werecontrastedincomparablysunlit,openen- the two well-lit habitats (open farm plot and forest
vironments (mean of roughly 1.7 yr for upper canopy canopy)hadshorterleaflifespanthantreesinthesmall
forest trees or open-grown saplings) (Table 2). gaps,withtreesinthedeeplyshadedforestunderstory
Among the three primary communities, the average having the longest life span of all habitats.
leaf life span in sunlit microhabitats ranged from 2.5 Contrastsamongspecieswithinsites.—Amongspe-
yr in Caatinga, to 1.9 yr in Bana, and 1.6 yr in Tierra ciesgrowinginacommonenvironment,thosewiththe
Firme,withsubstantialvariationamongspeciesineach shortestmedianleaflifespantendedtoshowagreater
community (Table 2). Contrasts of the two tall forest riskofmortalityforleavesateveryleaflifestage(the
types (Caatinga and TierraFirme)provideevidenceto survivorship curves continued to diverge over time)
test Hypothesis 1 (regarding differences in leaf life (Fig. 1). This was generally true, although there were
span among forest communities differing in resource exceptions.
availability).Caatingaspecieshadlongerleaflifespan, Forinstance,intheBanacommunity21%ofallAs-
on average, than Tierra Firme species in each of four pidospermumleavesdied by the ageof270d(Fig.1),
comparably lit habitats (Table 2): 2.3 vs. 1.8 yr when similar to the 20% mortality by 270 d shown by Re-
planted in an open farm plot; 2.5 vs. 1.6 yr in well-lit tiniphyllum. Yet average leaf life span in Aspidosper-
matureforestcanopy;3.4vs.1.9yringap;and4.2vs. mum was 2.4 yr (888 d) vs. 1.3 yr (474 d) for Reti-
3.2 yr in natural understory. In analysis of variance, niphyllum (Table 2), because the mortality rates for
these differences were significant for both community older leaves were greater for the latter species. Two
type (P (cid:44) 0.0001) and habitat (P (cid:44) 0.0001), and the other Bana species (Protium sp. and Rhodognaphal-
differences between communities was not affected by opsis) with median leaf life span similar to Aspidos-
habitat (P (cid:53) 0.49). These contrasts of Caatinga (av- permum had substantially lower cumulative mortality
erage leaf N (cid:53) 1.02 (cid:54) 0.012%; mean (cid:54) 1 SE) and (7% and 6%, respectively) by leaf age of 270 d than
TierraFirmespecies(averageleafN(cid:53)1.65(cid:54)0.12%; Aspidospermum. However, the monthly mortality rate
P (cid:44) 0.005) are consistent with the hypothesis that for remained low in Aspidospermum until leaves were 5
communitieswithsimilarlightenvironments,andcon- yrold(shallowslope,Fig.1),whereasforProtiumand
10 PETERB.REICHET AL. EcologicalMonographs
Vol.74,No.1
FIG.3. Mortalitycurves(Kaplan-Meiersurvivorshipcurve)forasubsetofspeciesacrosslightenvironments.Shownare
twospecieseachfromTierraFirmeandCaatingaforests,eachcomparedinmaturetallforestcanopy(open),forestgap,and
understoryenvironments.All curvesweresignificantlydifferent.
Rhodognaphalopsis, the monthly mortality rate rose postagriculturalsites;inthesecases,leaflifespanwas
more quickly with leaf age (steeper slope). Hence the twiceasgreatinthemoreshadedindividuals(Table2,
survivorshipcurvescrossforAspidospermumandthese P (cid:44) 0.001). Large differences in leaf life span were
other species (Fig. 1 showing all five Bana species also observed for the mid-to-intolerant Bellucia com-
curves). By leaf age of 4 yr old, 15% of all Aspidos- paredindifferentlightenvironments.Belluciahadme-
permum leaves were still alive, contrasted with 2% of dian leaf life span of 0.40–0.48 yr in three different
Protium (Bana) and none of Rhodognaphalopsis. recentlyabandonedfarmfields(dataotherwisepooled),
A similar pattern (including a crossover in survi- andleaflifespanof2.0yrgrowingintheTierraFirme
vorship curves) occurred in the Caatinga understory understory.
forMicrandra,Micropholis,andProtium,allofwhich Additionally,inrelativelycomparablelightenviron-
hadsimilaraverageleaflifespan((cid:59)4.5yr).Micrandra ments, species had similar leaf life spans (Table 2),
had greater early mortality (compare species at 250– regardless of other potential sources of variation. Six
500 d leaf age) than all four other Caatinga species, species common to tall forests had similar leaf life
but its mortality rates remained low for many years. spans, on average, as young trees planted in the open
Hence, 11% of all Micrandra leaves were alive atage (2.1 yr, average of the six species) as in natural sunlit
7.6 yr when only negligible percentages of leaves of upper canopy positions for older trees (2.0 yr).
other Caatinga species were alive.
Variation among trees and with seasons and years
Contrasts within species among sites and microsi-
tes.—For 10 species, leaf demography was tracked in How variable is leaf production and mortality from
differentlightenvironments(Table2).Inallcases,leaf treetotree,seasontoseasonoryeartoyear?Different
lifespanwassignificantly(P(cid:44)0.001)longerinshaded individual trees of a species growing in the same en-
than open-grown sites, either for naturally established vironment tended to show different temporal patterns
or planted individuals. For naturally established trees of leaf production or mortality (e.g., Fig. 4). For all
within Tierra Firme and Caatinga, census data were species, there was considerable variability in median
comparedforlargetreeswithsunlitcrownsvs.smaller leaf life span among trees within a site (Fig. 5), es-
trees in gaps and in the understory within the same peciallyfortreeswithlowleafpopulations.Giventhat
forest stand. In every such case, there were large sig- only one or two branches were surveyed per tree, and
nificant differences in leaf survival among light envi- that there is high variability in light microsites in the
ronments (Table 2, Fig. 3), with 50–100% greaterleaf understoryandevenintheuppercanopy(e.g.,Tjoelker
life span in more shaded environments. Three Tierra et al. 1995), it is difficult to assess how much of the
Firmespecieswerealsoplantedinopenandinshaded tree-to-tree variability shown in Fig. 5 is due to dif-
February2004 TROPICALRAINFORESTLEAFCENSUS 11
FIG.4. MonthlyleafmortalityandproductionratesforfourRetiniphyllumtreesovera5-yeartimeframe.Foreachcensus
interval, mortality and production rateswerecalculatedforthepopulationofleavesattheendoftheintervalinrelationto
the populationof leavesatthe beginningof the interval.
ferences in branch microenvironment, how much to icantdifferences(P(cid:44)0.001)insurvivorshipforleaves
genotypic variation among trees, and how much due produced in different months and years (Fig. 6b)orin
to branches with low leaf sample sizes. differentseasons,poolingacrossyears(Fig.6c).How-
ForfourrandomlyselectedRetiniphyllumtreeswith ever, the differences were relatively small. For exam-
relatively large sample sizes, different trees had gen- ple,leavesproducedinAugust–OctoberorNovember–
erally similar overall survivorship curves (e.g., Fig. January,beforeorduringthedriesttimeofyear(Table
6a), and similar median leaf life span and leaf age 1), respectively, had slightly shorter median life span
populationstructure.ForallRetiniphyllumleaves(n(cid:53) (1.19 and 1.23 yr, respectively) than those produced
4571) pooled across trees (n (cid:53) 31), there weresignif- before(February–April)orduring(May–July)thewet-
12 PETERB.REICHET AL. EcologicalMonographs
Vol.74,No.1
FIG. 5. The frequency distribution among trees of median leaf life span (pertree)forsevenspecies,threeofwhichare
shown in contrastinglightmicroenvironments.
test time of year (1.28 to 1.32 yr, respectively). Dif- lum trees, survival for cohorts produced during the
ferencesinsurvivalovertimeamongleafcohortspro- same seasonal period, but in different years, varied as
ducedindifferenttimesofyearwereevensmaller,and much as among seasons within years (Fig. 6b). For
not significant (P (cid:46) 0.05), when the comparison was individual trees examined over time (e.g., Fig. 4), we
restricted to a single tree (Fig. 6d). For all Retiniphyl- see no evidence of climate-related variation in pro-